X-Git-Url: http://info.iut-bm.univ-fcomte.fr/pub/gitweb/simgrid.git/blobdiff_plain/c5ad8ca1a68bbaa9152471c8d0eeb99d762f0d86..23f5bf52cc58cf837840f231cf638e9fcd45097b:/src/simix/smx_host.c diff --git a/src/simix/smx_host.c b/src/simix/smx_host.c index e87041bc51..2308f27d4b 100644 --- a/src/simix/smx_host.c +++ b/src/simix/smx_host.c @@ -1,4 +1,4 @@ -/* Copyright (c) 2007, 2008, 2009, 2010. The SimGrid Team. +/* Copyright (c) 2007-2012. The SimGrid Team. * All rights reserved. */ /* This program is free software; you can redistribute it and/or modify it @@ -13,7 +13,6 @@ XBT_LOG_NEW_DEFAULT_SUBCATEGORY(simix_host, simix, "Logging specific to SIMIX (hosts)"); - static void SIMIX_execution_finish(smx_action_t action); /** @@ -67,7 +66,7 @@ void SIMIX_host_destroy(void *h) SIMIX_display_process_status(); THROWF(arg_error, 0, "%s", msg); } - + xbt_dynar_free(&host->auto_restart_processes); xbt_swag_free(host->process_list); /* Clean host structure */ @@ -90,7 +89,7 @@ xbt_dict_t SIMIX_host_get_dict(void) void **host = NULL; xbt_lib_foreach(host_lib, cursor, name, host){ - if(host[SIMIX_HOST_LEVEL]) + if(host[SIMIX_HOST_LEVEL]) xbt_dict_set(host_dict,name,host[SIMIX_HOST_LEVEL], NULL); } return host_dict; @@ -176,6 +175,107 @@ void* SIMIX_host_get_data(smx_host_t host) return host->data; } +void _SIMIX_host_free_process_arg(void *); +void _SIMIX_host_free_process_arg(void *data) +{ + smx_process_arg_t arg = *(void**)data; + xbt_free(arg->name); + xbt_free(arg); +} +/** + * \brief Add a process to the list of the processes that the host will restart when it comes back + * This function add a process to the list of the processes that will be restarted when the host comes + * back. It is expected that this function is called when the host is down. + * The processes will only be restarted once, meaning that you will have to register the process + * again to restart the process again. + */ +void SIMIX_host_add_auto_restart_process(smx_host_t host, + const char *name, + xbt_main_func_t code, + void *data, + const char *hostname, + double kill_time, + int argc, char **argv, + xbt_dict_t properties, + int auto_restart) +{ + if (!host->auto_restart_processes) { + host->auto_restart_processes = xbt_dynar_new(sizeof(smx_process_arg_t),_SIMIX_host_free_process_arg); + } + smx_process_arg_t arg = xbt_new(s_smx_process_arg_t,1); + arg->name = xbt_strdup(name); + arg->code = code; + arg->data = data; + arg->hostname = hostname; + arg->kill_time = kill_time; + arg->argc = argc; + + arg->argv = xbt_new(char*,argc + 1); + + int i; + for (i = 0; i < argc; i++) { + arg->argv[i] = xbt_strdup(argv[i]); + } + arg->argv[argc] = NULL; + + arg->properties = properties; + arg->auto_restart = auto_restart; + + if( SIMIX_host_get_state(host) == SURF_RESOURCE_OFF + && !xbt_dict_get_or_null(watched_hosts_lib,host->name)){ + xbt_dict_set(watched_hosts_lib,host->name,host,NULL); + XBT_DEBUG("Have push host %s to watched_hosts_lib because state == SURF_RESOURCE_OFF",host->name); + } + xbt_dynar_push_as(host->auto_restart_processes,smx_process_arg_t,arg); +} +/** + * \brief Restart the list of processes that have been registered to the host + */ +void SIMIX_host_restart_processes(smx_host_t host) +{ + unsigned int cpt; + smx_process_arg_t arg; + xbt_dynar_foreach(host->auto_restart_processes,cpt,arg) { + + smx_process_t process; + + XBT_DEBUG("Restarting Process %s(%s) right now", arg->argv[0], arg->hostname); + if (simix_global->create_process_function) { + simix_global->create_process_function(&process, + arg->argv[0], + arg->code, + NULL, + arg->hostname, + arg->kill_time, + arg->argc, + arg->argv, + arg->properties, + arg->auto_restart); + } + else { + simcall_process_create(&process, + arg->argv[0], + arg->code, + NULL, + arg->hostname, + arg->kill_time, + arg->argc, + arg->argv, + arg->properties, + arg->auto_restart); + + } + } + xbt_dynar_reset(host->auto_restart_processes); +} + +void SIMIX_host_autorestart(smx_host_t host) +{ + if(simix_global->autorestart) + simix_global->autorestart(host); + else + xbt_die("No function for simix_global->autorestart"); +} void SIMIX_host_set_data(smx_host_t host, void *data) { @@ -201,10 +301,10 @@ smx_action_t SIMIX_host_execute(const char *name, smx_host_t host, #endif /* set surf's action */ - if (!MC_IS_ENABLED) { + if (!MC_is_active()) { action->execution.surf_exec = surf_workstation_model->extension.workstation.execute(host->host, - computation_amount); + computation_amount); surf_workstation_model->action_data_set(action->execution.surf_exec, action); surf_workstation_model->set_priority(action->execution.surf_exec, priority); } @@ -239,11 +339,11 @@ smx_action_t SIMIX_host_parallel_execute( const char *name, workstation_list[i] = host_list[i]->host; /* set surf's action */ - if (!MC_IS_ENABLED) { + if (!MC_is_active()) { action->execution.surf_exec = surf_workstation_model->extension.workstation. execute_parallel_task(host_nb, workstation_list, computation_amount, - communication_amount, amount, rate); + communication_amount, rate); surf_workstation_model->action_data_set(action->execution.surf_exec, action); } @@ -256,13 +356,11 @@ void SIMIX_host_execution_destroy(smx_action_t action) { XBT_DEBUG("Destroy action %p", action); - xbt_free(action->name); - if (action->execution.surf_exec) { surf_workstation_model->action_unref(action->execution.surf_exec); action->execution.surf_exec = NULL; } - + xbt_free(action->name); xbt_mallocator_release(simix_global->action_mallocator, action); } @@ -299,14 +397,14 @@ void SIMIX_pre_host_execution_wait(smx_simcall_t simcall) { smx_action_t action = simcall->host_execution_wait.execution; - XBT_DEBUG("Wait for execution of action %p, state %d", action, action->state); + XBT_DEBUG("Wait for execution of action %p, state %d", action, (int)action->state); /* Associate this simcall to the action */ xbt_fifo_push(action->simcalls, simcall); simcall->issuer->waiting_action = action; /* set surf's action */ - if (MC_IS_ENABLED) { + if (MC_is_active()) { action->state = SIMIX_DONE; SIMIX_execution_finish(action); return; @@ -326,12 +424,12 @@ void SIMIX_host_execution_suspend(smx_action_t action) void SIMIX_host_execution_resume(smx_action_t action) { if(action->execution.surf_exec) - surf_workstation_model->suspend(action->execution.surf_exec); + surf_workstation_model->resume(action->execution.surf_exec); } void SIMIX_execution_finish(smx_action_t action) { - volatile xbt_fifo_item_t item; + xbt_fifo_item_t item; smx_simcall_t simcall; xbt_fifo_foreach(action->simcalls, item, simcall, smx_simcall_t) { @@ -340,34 +438,31 @@ void SIMIX_execution_finish(smx_action_t action) case SIMIX_DONE: /* do nothing, action done */ - XBT_DEBUG("SIMIX_execution_finished: execution successful"); + XBT_DEBUG("SIMIX_execution_finished: execution successful"); break; case SIMIX_FAILED: XBT_DEBUG("SIMIX_execution_finished: host '%s' failed", simcall->issuer->smx_host->name); - TRY { - THROWF(host_error, 0, "Host failed"); - } - CATCH(simcall->issuer->running_ctx->exception) { - simcall->issuer->doexception = 1; - } - break; + simcall->issuer->context->iwannadie = 1; + //SMX_EXCEPTION(simcall->issuer, host_error, 0, "Host failed"); + break; case SIMIX_CANCELED: XBT_DEBUG("SIMIX_execution_finished: execution canceled"); - TRY { - THROWF(cancel_error, 0, "Canceled"); - } - CATCH(simcall->issuer->running_ctx->exception) { - simcall->issuer->doexception = 1; - } - break; + SMX_EXCEPTION(simcall->issuer, cancel_error, 0, "Canceled"); + break; default: xbt_die("Internal error in SIMIX_execution_finish: unexpected action state %d", - action->state); + (int)action->state); + } + /* check if the host is down */ + if (surf_workstation_model->extension. + workstation.get_state(simcall->issuer->smx_host->host) != SURF_RESOURCE_ON) { + simcall->issuer->context->iwannadie = 1; } - simcall->issuer->waiting_action = NULL; + + simcall->issuer->waiting_action = NULL; simcall->host_execution_wait.result = action->state; SIMIX_simcall_answer(simcall); } @@ -378,14 +473,19 @@ void SIMIX_execution_finish(smx_action_t action) void SIMIX_post_host_execute(smx_action_t action) { - /* FIXME: check if the host running the action failed or not*/ - /*if(surf_workstation_model->extension.workstation.get_state(action->host->host))*/ - - /* If the host running the action didn't fail, then the action was canceled */ - if (surf_workstation_model->action_state_get(action->execution.surf_exec) == SURF_ACTION_FAILED) - action->state = SIMIX_CANCELED; - else - action->state = SIMIX_DONE; + if (action->type == SIMIX_ACTION_EXECUTE && /* FIMXE: handle resource failure + * for parallel tasks too */ + surf_workstation_model->extension.workstation.get_state(action->execution.host->host) == SURF_RESOURCE_OFF) { + /* If the host running the action failed, notice it so that the asking + * process can be killed if it runs on that host itself */ + action->state = SIMIX_FAILED; + } else if (surf_workstation_model->action_state_get(action->execution.surf_exec) == SURF_ACTION_FAILED) { + /* If the host running the action didn't fail, then the action was + * canceled */ + action->state = SIMIX_CANCELED; + } else { + action->state = SIMIX_DONE; + } if (action->execution.surf_exec) { surf_workstation_model->action_unref(action->execution.surf_exec);